Good and well explained course !

Thank you! I enjoyed the course

Excellent hands-on learning!

very clearly explained

Really a good course

Good Course Content

Intuitive hands-on.

awesome course

Great course!!

Great course !

great course

Really clear!

Very Good..!!

Good Course

Love it....

Nice Info

Awesome

Good!!

good

This is a good introduction to Tensorflow. Like all Coursera courses I've experienced to date, there were plusses and minuses.

The good side of each of these courses: * The courses cover the main concepts (building models, limitations, challenges, etc). They covered activation functions, Convolutions, width and depth of models, Gradient Descent and learning rate issues.

* The lessons don't oversimplify, but give you the tools you need to explore further on your own if you wish.

* Replies to my forum questions were actually surprisingly quickly answered. I was expecting the forums to be filled with months-old unanswered questions.

* Being able to replay videos was invaluable.

The less-good side:

* The exams are token, often multiple choice with unlimited retries. That is fine.

* The lessons are often replete with misspellings, grammar errors and ambiguous quiz questions.

* Sometimes, due to the stochastic nature of ML models, the errors/mispredictions differ between the Grading Rubrics and legitimately obtained results.

Would I do it again? My answer is this- I feel for six courses I have the equivalent of one junior-level semester survey course's worth of information and experience. However I was able to do it on my own time starting immediately, at my own pace, replaying the lectures at will and all for a tiny fraction of the cost and time of a college course. I do believe I have a starting point to pursue more advanced topics and for that I believe it was well worth it.

I felt the labs failed to illustrate the reasons why we were learning the concepts. They did not use examples that would have shown how the tools would have value with real projects. For example: Building Deep Learning Models with TensorFlow/ML0120EN-3.1-Reveiw-LSTM-basics only illustrated some very limited concepts. Instead of an example that had value, it was just variables and numbers. The same is true for labs_ML0120EN-3.2-Review-LSTM-LanguageModelling_with_results.ipynb. It did not actually model anything of relevance. It had a section on # Define the gradient clipping threshold, without explaining why that is important. This was true with most of the exercises. When I compare that with a course like Introduction to Deep Learning & Neural Networks with Keras the differences are vast. Keras is a simple interface and all the examples were clear and had real world applications for business. Not so with this course. A huge disappointment, and a terrible waste of time.

This is a great course and a great instructor. I also loved his course on Machine Learning with Python. My major criticism, relevant also for the course on Keras in the AI Engineering program, is that the lectures and labs are not updated to the new versions of packages. The new versions of Tensorflow, Tensorflow2.0, were changed significantly relative to the version used here. Moreover, Keras in now TensorFlow's official high-level API, which means that the code learned in these courses cannot be used for new data without implementing the new syntax of these libraries. I hope IBM will update the learning material more frequently so these wonderful courses will keep being relevant.

The instructor Saeed Aghabozorgi did an excellent job in explaining the concepts in a way everything can be understood easily. However, I still think 5 weeks is not enough for this course, given TensorFlow is more difficult to learn than PyTorch. The basics could be covered in more detail, including the tf.get_variable(), tf.gradient(), calculating gradients and other functions that were used. There could be a lecture for Linear Regression and Logistic Regression and these 2 could be moved to a separate week instead. Also, please upgrade the code to work on TensorFlow 2.1. The current code designed for TensorFlow 1.8 didn't work especially the part where datasets are to be loaded.

Four stars because some of the labs (and none of the lectures) have not been brought up to the current version of TensorFlow. There are significant differences between 1.x and 2.x, especially in the paralell processing. I don't expect a course to send me on wild goose chases across the internet having to bring their examples up to current versions. I guess you get what you pay for, no surprise that Big Blue isn't current.